File size: 2,543 Bytes
7b12cd6
46f3e04
 
 
444992d
 
 
46f3e04
 
444992d
 
46f3e04
 
 
 
 
444992d
46f3e04
 
 
 
 
 
 
444992d
46f3e04
444992d
7b12cd6
 
46f3e04
 
7b12cd6
7435aa1
7b12cd6
46f3e04
 
 
 
7b12cd6
46f3e04
7b12cd6
ce17a5f
 
 
 
 
 
 
 
 
 
 
 
1b2c72d
7b12cd6
46f3e04
7b12cd6
46f3e04
7b12cd6
46f3e04
7b12cd6
46f3e04
7b12cd6
46f3e04
7b12cd6
46f3e04
7b12cd6
46f3e04
 
 
 
 
 
 
 
 
 
7b12cd6
46f3e04
7b12cd6
46f3e04
 
 
 
 
 
7b12cd6
 
46f3e04
7b12cd6
46f3e04
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
---
base_model: openai/whisper-small
datasets:
- mozilla-foundation/common_voice_11_0
language:
- yo
license: apache-2.0
metrics:
- wer
tags:
- generated_from_trainer
model-index:
- name: Whisper Small Yo - Bola Ologundudu
  results:
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: Common Voice 11.0
      type: mozilla-foundation/common_voice_11_0
      config: yo
      split: None
      args: 'config: yo, split: test'
    metrics:
    - type: wer
      value: 70.61345018098686
      name: Wer
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# Whisper Small Yoruba - Bola Ologundudu

This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the Common Voice 11.0 dataset.
It achieves the following results on the evaluation set:
- Loss: 1.2225
- Wer: 70.6135

## Model description

>>> from transformers import pipeline
>>> import torch

>>> modelName="ajibs75/whisper-small-yoruba"
>>> device = 0 if torch.cuda.is_available() else "cpu"
>>> pipe = pipeline(task="automatic-speech-recognition",model=modelName,chunk_length_s=30,device=device,)
>>> pipe.model.config.forced_decoder_ids = pipe.tokenizer.get_decoder_prompt_ids(language="yo", task="transcribe")

>>> audio = "sample.mp3"
>>> text = pipe(audio)
>>> transacribed_audio = text["text"] 
>>> print(transacribed_audio)


## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 4000
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch   | Step | Validation Loss | Wer     |
|:-------------:|:-------:|:----:|:---------------:|:-------:|
| 0.066         | 7.6923  | 1000 | 0.8962          | 74.0141 |
| 0.004         | 15.3846 | 2000 | 1.1411          | 71.6613 |
| 0.0004        | 23.0769 | 3000 | 1.1959          | 70.6516 |
| 0.0003        | 30.7692 | 4000 | 1.2225          | 70.6135 |


### Framework versions

- Transformers 4.42.4
- Pytorch 2.1.0+cu118
- Datasets 2.20.0
- Tokenizers 0.19.1